LlamaIndex's core product, LlamaParse, leverages Vision Language Models (VLMs) and task-specific agentic pipelines to deliver layout-aware document parsing. It routes document content — text, tables, charts, handwriting — to specialized expert agents, then applies recursive auto-correction loops to maximize accuracy. The platform exposes APIs for parse, extract (schema-based LLM extraction), split, classify, index, and retrieval operations, making it suitable for building full RAG pipelines and multi-step document agents. An open-source sibling, LiteParse (npm install @llamaindex/liteparse), enables fast local parsing with no cloud dependencies or LLM token usage, outputting bounding-box-annotated structured text from PDFs, Office docs, and images.